NETANYA, Israel Mixed signal ASIC specialist ChipX Inc. has implemented a complex, 10 million gate design for pattern recognition, using Recognetics' neural network technology.
The device, dubbed the CM1K, is a neural network chip featuring 1024 neurons working in parallel that can be daisy-chained to other CM1K devices to increase network size and functionality.
The chip is targeted at smart sensors and cameras and can classify patterns at high speed while coping with ill-defined data and unknown events, and adapting to changes of contexts and working conditions. It is expected to find applications in markets such as security, data mining, imaging, robotics, manufacturing and others.
Recognition time of a 256-byte vector is achieved in less than 10us while operating at 27MHz, keeping power consumption of the device very low, according to ChipX. The device uses 3.3V and 1.2V supplies and dissipates only 300mW. CM1K can extract 1D signatures from 2D camera data.
Wo Lin, President of Recognetics, commented: "The team worked diligently to achieve the CM1K's performance, power and cost objectives and we are delighted with the results."
"We are excited about Recognetics' breakthrough technology and its potential use in a variety of pattern recognition applications," said Elie Massabki, vice president of marketing at ChipX.
The CM1K starts sampling this week and a variety of demonstration and implementation platforms are available from Recognetics directly.
Headquartered in Santa Clara, Calif., ChipX is a privately held group, with a Research and Development subsidiary in Israel. Investors include Elron Electronic Industries, Ltd., VantagePoint Venture Partners, Wasserstein Venture Capital, UMC and Needham Capital Partners.
Recognetics is a leading provider for fully parallel, high performance pattern recognition semiconductors based on the CogniMem neural network patented technology.